From: Shaohua Li <shli@kernel.org>
To: linux-raid@vger.kernel.org, linux-kernel@vger.kernel.org
Cc: neilb@suse.de, djbw@fb.com, tj@kernel.org
Subject: [patch 3/3 v2] raid5: only wakeup necessary threads
Date: Mon, 12 Aug 2013 10:18:06 +0800 [thread overview]
Message-ID: <20130812022017.500088923@kernel.org> (raw)
In-Reply-To: 20130812021803.325887805@kernel.org
[-- Attachment #1: raid5-intelligent-wakeup.patch --]
[-- Type: text/plain, Size: 6199 bytes --]
If there are no enough stripes to handle, we'd better not always queue all
available work_structs. If one worker can only handle small or even none
stripes, it will impact request merge and create lock contention.
With this patch, the number of work_struct running will depend on pending
stripes number. Note: some statistics info used in the patch are accessed without
locking protection. This should doesn't matter, we just try best to avoid queue
unnecessary work_struct.
Signed-off-by: Shaohua Li <shli@fusionio.com>
---
drivers/md/raid5.c | 50 ++++++++++++++++++++++++++++++++++++++++++++------
drivers/md/raid5.h | 4 ++++
2 files changed, 48 insertions(+), 6 deletions(-)
Index: linux/drivers/md/raid5.c
===================================================================
--- linux.orig/drivers/md/raid5.c 2013-08-09 09:50:25.754255596 +0800
+++ linux/drivers/md/raid5.c 2013-08-09 10:03:45.016208049 +0800
@@ -77,6 +77,7 @@ static struct workqueue_struct *raid5_wq
#define BYPASS_THRESHOLD 1
#define NR_HASH (PAGE_SIZE / sizeof(struct hlist_head))
#define HASH_MASK (NR_HASH - 1)
+#define MAX_STRIPE_BATCH 8
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
{
@@ -209,6 +210,7 @@ static void raid5_wakeup_stripe_thread(s
{
struct r5conf *conf = sh->raid_conf;
struct r5worker_group *group;
+ int thread_cnt;
int i;
if (conf->worker_cnt_per_group == 0) {
@@ -218,8 +220,26 @@ static void raid5_wakeup_stripe_thread(s
group = conf->worker_groups + cpu_to_group(sh->cpu);
- for (i = 0; i < conf->worker_cnt_per_group; i++)
- queue_work_on(sh->cpu, raid5_wq, &group->workers[i].work);
+ group->workers[0].working = true;
+ /* at least one worker should run to avoid race */
+ queue_work_on(sh->cpu, raid5_wq, &group->workers[0].work);
+
+ thread_cnt = group->stripes_cnt / MAX_STRIPE_BATCH - 1;
+ /* wakeup more workers */
+ for (i = 1; i < conf->worker_cnt_per_group && thread_cnt > 0; i++) {
+ if (group->workers[i].working == false) {
+ group->workers[i].working = true;
+ queue_work_on(sh->cpu, raid5_wq,
+ &group->workers[i].work);
+ thread_cnt--;
+ } else if (group->workers[i].working_cnt <=
+ MAX_STRIPE_BATCH / 2)
+ /*
+ * If a worker has no enough stripes handling, assume
+ * it will fetch more stripes soon.
+ */
+ thread_cnt--;
+ }
}
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
@@ -248,6 +268,8 @@ static void do_release_stripe(struct r5c
struct r5worker_group *group;
group = conf->worker_groups + cpu_to_group(cpu);
list_add_tail(&sh->lru, &group->handle_list);
+ group->stripes_cnt++;
+ sh->group = group;
}
raid5_wakeup_stripe_thread(sh);
return;
@@ -573,6 +595,10 @@ get_active_stripe(struct r5conf *conf, s
!test_bit(STRIPE_EXPANDING, &sh->state))
BUG();
list_del_init(&sh->lru);
+ if (sh->group) {
+ sh->group->stripes_cnt--;
+ sh->group = NULL;
+ }
}
}
} while (sh == NULL);
@@ -4143,6 +4169,7 @@ static struct stripe_head *__get_priorit
{
struct stripe_head *sh = NULL, *tmp;
struct list_head *handle_list = NULL;
+ struct r5worker_group *wg = NULL;
if (conf->worker_cnt_per_group == 0) {
handle_list = &conf->handle_list;
@@ -4150,12 +4177,14 @@ static struct stripe_head *__get_priorit
handle_list = NULL;
} else if (group != ANY_GROUP) {
handle_list = &conf->worker_groups[group].handle_list;
+ wg = &conf->worker_groups[group];
if (list_empty(handle_list))
handle_list = NULL;
} else {
int i;
for (i = 0; i < conf->group_cnt; i++) {
handle_list = &conf->worker_groups[i].handle_list;
+ wg = &conf->worker_groups[i];
if (!list_empty(handle_list))
break;
}
@@ -4204,11 +4233,16 @@ static struct stripe_head *__get_priorit
if (conf->bypass_count < 0)
conf->bypass_count = 0;
}
+ wg = NULL;
}
if (!sh)
return NULL;
+ if (wg) {
+ wg->stripes_cnt--;
+ sh->group = NULL;
+ }
list_del_init(&sh->lru);
atomic_inc(&sh->count);
BUG_ON(atomic_read(&sh->count) != 1);
@@ -4906,8 +4940,8 @@ static int retry_aligned_read(struct r5
return handled;
}
-#define MAX_STRIPE_BATCH 8
-static int handle_active_stripes(struct r5conf *conf, int group)
+static int handle_active_stripes(struct r5conf *conf, int group,
+ struct r5worker *worker)
{
struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
int i, batch_size = 0;
@@ -4916,6 +4950,9 @@ static int handle_active_stripes(struct
(sh = __get_priority_stripe(conf, group)) != NULL)
batch[batch_size++] = sh;
+ if (worker)
+ worker->working_cnt = batch_size;
+
if (batch_size == 0)
return batch_size;
spin_unlock_irq(&conf->device_lock);
@@ -4950,11 +4987,12 @@ static void raid5_do_work(struct work_st
released = release_stripe_list(conf);
- batch_size = handle_active_stripes(conf, group_id);
+ batch_size = handle_active_stripes(conf, group_id, worker);
if (!batch_size && !released)
break;
handled += batch_size;
}
+ worker->working = false;
pr_debug("%d stripes handled\n", handled);
spin_unlock_irq(&conf->device_lock);
@@ -5012,7 +5050,7 @@ static void raid5d(struct md_thread *thr
handled++;
}
- batch_size = handle_active_stripes(conf, ANY_GROUP);
+ batch_size = handle_active_stripes(conf, ANY_GROUP, NULL);
if (!batch_size && !released)
break;
handled += batch_size;
Index: linux/drivers/md/raid5.h
===================================================================
--- linux.orig/drivers/md/raid5.h 2013-08-09 09:42:47.076021908 +0800
+++ linux/drivers/md/raid5.h 2013-08-09 09:54:27.927211264 +0800
@@ -213,6 +213,7 @@ struct stripe_head {
enum reconstruct_states reconstruct_state;
spinlock_t stripe_lock;
int cpu;
+ struct r5worker_group *group;
/**
* struct stripe_operations
* @target - STRIPE_OP_COMPUTE_BLK target
@@ -369,12 +370,15 @@ struct disk_info {
struct r5worker {
struct work_struct work;
struct r5worker_group *group;
+ int working_cnt:8;
+ bool working;
};
struct r5worker_group {
struct list_head handle_list;
struct r5conf *conf;
struct r5worker *workers;
+ int stripes_cnt;
};
struct r5conf {
next prev parent reply other threads:[~2013-08-12 2:18 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-08-12 2:18 [patch 0/3 v2] raid5: make stripe handling multi-threading Shaohua Li
2013-08-12 2:18 ` [patch 1/3 v2] raid5: offload stripe handle to workqueue Shaohua Li
2013-08-12 2:18 ` [patch 2/3 v2] raid5: sysfs entry to control worker thread number Shaohua Li
2013-08-12 2:18 ` Shaohua Li [this message]
2013-08-27 7:34 ` [patch 0/3 v2] raid5: make stripe handling multi-threading NeilBrown
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130812022017.500088923@kernel.org \
--to=shli@kernel.org \
--cc=djbw@fb.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-raid@vger.kernel.org \
--cc=neilb@suse.de \
--cc=tj@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).